Obstacle distribution simulation method and device based on multiple models, and storage medium

Abstract

An obstacle distribution simulation method, device and terminal based on multiple models. The method can include: acquiring a point cloud, the point cloud including a plurality of obstacles labeled with real labeling data; extracting the real labeling data of the obstacles, and training a plurality of neural network models based on the real labeling data of the obstacles; extracting unlabeled data in the point cloud, inputting the unlabeled data into the neural network models, and outputting a plurality of prediction results. The plurality of prediction results can include a plurality of simulated obstacles with attribute data; selecting at least one simulated obstacle based on the plurality of prediction results; and inputting the attribute data of the selected simulated obstacle into the neural network models to obtain position coordinates of the simulated obstacle, and further obtain a position distribution of the simulated obstacle.

What is claimed is: 
     
       1. An obstacle distribution simulation method based on multiple models, comprising:
 acquiring a point cloud based on a plurality of frames, the point cloud comprising a plurality of obstacles labeled with real labeling data; 
 extracting the real labeling data of the plurality of obstacles, and training a plurality of neural network models based on the real labeling data of the plurality of obstacles; 
 extracting unlabeled data of unlabeled obstacles from the point cloud, inputting the unlabeled data into the plurality of neural network models, and outputting, for the unlabeled data, a plurality of prediction results obtained by the plurality of neural network models, wherein the plurality of prediction results comprise a plurality of simulated obstacles with attribute data; 
 selecting a simulated obstacle based on the plurality of prediction results; and 
 inputting the attribute data of the selected simulated obstacle into the plurality of neural network models to obtain position coordinates of the simulated obstacle using the plurality of neural network models, and further obtain a position distribution of the simulated obstacle using the plurality of neural network models. 
 
     
     
       2. The method of  claim 1 , further comprising:
 outputting confidences corresponding to the respective prediction results; and 
 determining whether the confidences are greater than a threshold, and reserving a prediction result with a confidence greater than the threshold. 
 
     
     
       3. The method of  claim 1 , wherein selecting at least one simulated obstacle based on the plurality of prediction results comprises:
 determining whether the plurality of prediction results comprise a common simulated obstacle, and selecting the common simulated obstacle when the plurality of prediction results comprise the common simulated obstacle. 
 
     
     
       4. The method of  claim 1 , wherein inputting the attribute data of the selected simulated obstacle into the plurality of neural network models to obtain position coordinates of the simulated obstacle comprises:
 inputting the attribute data of the selected simulated obstacle into the plurality of neural network models to obtain a plurality of boundary boxes of the simulated obstacle; 
 obtaining a length and a width of the each of the plurality of boundary boxes of the simulated obstacle; 
 calculating an average length value and an average width value based on the lengths and widths of the plurality of boundary boxes; and 
 calculating center coordinates of an average boundary box based on the average length value and the average width value such that the center coordinates are represented as position coordinates of the simulated obstacle. 
 
     
     
       5. An obstacle distribution simulation device based on multiple models, the device comprising:
 one or more processors; and 
 a storage device configured to store one or more programs, that, when executed by the one or more processors, cause the one or more processors to:
 acquire a point cloud based on a plurality of frames, the point cloud comprising a plurality of obstacles labeled by real labeling data; 
 extract the real labeling data of the plurality of obstacles, and train a plurality of neural network models based on the real labeling data of the plurality of obstacles; 
 extract unlabeled data of unlabeled obstacles from the point cloud, input the unlabeled data into the plurality of neural network models, and output, for the unlabeled data, a plurality of prediction results obtained by the plurality of neural network models, wherein the plurality of prediction results comprise a plurality of simulated obstacles with attribute data; 
 select a simulated obstacle based on the plurality of prediction results; and 
 input the attribute data of the selected simulated obstacle into the plurality of neural network models to obtain position coordinates of the simulated obstacle using the plurality of neural network models, and to further obtain a position distribution of the simulated obstacle using the plurality of neural network models. 
 
 
     
     
       6. A non-transitory computer readable storage medium, in which a computer program is stored, wherein the program, when executed by a processor, causes the processor to implement the method of  claim 1 . 
     
     
       7. The device of  claim 5 , wherein the one or more programs, when executed by the one or more processors, cause the one or more processors further to:
 output a confidence corresponding to the each of the plurality of prediction results; 
 determine whether each confidence is greater than a threshold; 
 and retain each prediction result having a confidence greater than the threshold. 
 
     
     
       8. The device of  claim 5 , wherein the one or more programs, when executed by the one or more processors, cause the one or more processors further to:
 input the attribute data of the selected simulated obstacle into the plurality of neural network models to obtain a plurality of boundary boxes of the simulated obstacle; 
 obtain lengths and widths of the plurality of boundary boxes of the simulated obstacle; 
 calculate an average length value and an average width value based on the lengths and widths of the plurality of boundary boxes; and 
 calculate center coordinates of an average boundary box based on the average length value and the average width value, such that the center coordinates are represented as position coordinates of the simulated obstacle.